Lead AI Compiler Engineer

WHAT YOU DO AT AMD CHANGES EVERYTHING

Together, we advance your career.

Lead AI Compiler Engineer

ROLE OVERVIEW:

We are looking for an AI Compiler Engineer to join this high impact team working in the growing field of on-device AI inference acceleration as an individual contributor or as a technical lead in the AI group (AIG). As an AI Compiler Engineer, you will design and optimize AI compiler stack and tools that enable efficient execution of state-of-the-art open source as well as proprietary AI models such as LLMs, transformer models, etc., to AMD NPUs for on-device AI inference use-cases.. You will work on transforming high-level AI models into efficient, low-level code that can run on NPU. Your work will directly impact the performance, efficiency, and scalability of our AI solutions.

SCOPE OF WORK:

• Graph transformation

• Constant folding
• Operator fusion: Identify and implement performance optimization opportunities by reducing memory traffic through operator fusion at different memory hierarchy levels e.g., attention block
• Common subexpression elimination

• Problem partitioning and dataflow orchestration:

• Design of algorithms to optimally map given AI operation to the NPU comprising of an interconnected array of AI engines.

• Design and implementation of algorithms to orchestrate dataflow through multi-level memory hierarchy.

• Kernel Design and Development:

• Design and implement highly optimized C++/intrinsic based kernels for AI related operators.

• Develop vectorized code that leverages SIMD (Single Instruction, Multiple Data) and VLIW (Very Long Instruction Word) for optimal performance.

• Perform performance, program memory and accuracy tradeoffs.

• Testing and Validation:

• Develop CPU models for the ML operators in C++/ Python to validate accuracy.

• Write unit tests and integration tests to ensure correctness and reliability.

• Performance Profiling and Tuning:

• Profile and analyze the performance of model layers

• Identify performance/accuracy bottlenecks and alleviate those

• Documentation and Collaboration:

• Effective technical communication of day-to-day work and document design specs.

• Follow good coding practices, using version control system.

• Collaborate with cross-functional teams spanning over AI research, core architecture and software engineering.

REQUIRED SKILLS:

• Excellent C/C++ and Python coding skills

• Good understanding of SIMD, VLIW processor architecture.

• Experience with vectorized programming (SIMD).

• Thorough understanding of fixed and floating point arithmetic.

• Good understanding of various operators in state-of-the-art AI models.

• Knowledge of low-level hardware details (cache hierarchy, DMA programming).

• Excellent problem-solving skills especially on debug and a passion for on-device AI.

• Prefer candidates with past experience on AI compiler design.

ACADEMIC CREDENTIALS:

• BS/Masters/PhD degree in Computer Science, Electrical Engineering, or a related field.

#LI-NR1

Benefits offered are described: .

AMD does not accept unsolicited resumes from headhunters, recruitment agencies, or fee-based recruitment services. AMD and its subsidiaries are equal opportunity, inclusive employers and will consider all applicants without regard to age, ancestry, color, marital status, medical condition, mental or physical disability, national origin, race, religion, political and/or third-party affiliation, sex, pregnancy, sexual orientation, gender identity, military or veteran status, or any other characteristic protected by law. We encourage applications from all qualified candidates and will accommodate applicants needs under the respective laws throughout all stages of the recruitment and selection process.